Draft:Deep Tissue Optical Imaging of Near Cellular-Sized Features(DOLPHIN)

• Comment: In accordance with the Wikimedia Foundation's Terms of Use, I disclose that I have been paid by my employer for my contributions to this article. 152.3.43.59 (talk) 23:20, 21 April 2025 (UTC)

---

DOLPHIN: Deep-tissue Optical Imaging of Near Cellular-sized Features

Introduction

The DOLPHIN imagination system is used in conjunction with surveillance/detection devices, as well as a spectrometric apparatus for fluorescence hyperspectral imaging and hyperdiffuse imaging tables and lasers. When deep-fluorescence tissue imaging is conducted near cells with resolution near the cellular level, various modalities are highly scaled into a unified framework aided by the DOLPHIN construction platform. The exploration geometry configuration is marked by trans-illumination, where the InGaAs conduction-cooled camera on liquidfnitrogen captures images above the sample cage/box, while the 980 nm computer-controlled light-delivering laser provides illumination from below the sample. Alongside bright-field imaging for anatomical coregistration, a silicon-based camera portrays the general white field. To prevent background autofluorescence, a quartz platform fulfills the role and provides a workstation structure. The lensbox with a computer scanner mounted on a stage XY translatable stage moves along one axis to another, enabling features of a topograph machine to enable space mapping. The system switches between configurations, with band-pass filters for HDI and an equipped spectrometer with a monochromator for HSI, depending on the modes. Under advanced spectral analysis pipelines and dual-mode data acquisition, with biothermally enabled up to 8 cm tissue, DOLPHIN constructs 3D reconstructions of fluorescent samples, preparing submillimeter cuts and quasi-dissolving them over the middle regions. Data acquisition and processing are coordinated through custom LabVIEW and MATLAB software, with real-time imaging and spectral analysis. This includes optical methodology in conjunction with innovation regarding biomarker tagging, clinical applications, and future research pathways.

System Setup

The multifunctional DOLPHIN imaging system is integrated with a nitrogen cooled indium gallium arsenide (InGaAs) detector and a silicon technology camera. The system employs a trans-illumination configuration featuring both Hyperspectral Imaging (HSI) and Hyperdiffuse Imaging (HDI) modes. Illumination is provided by a computer-controlled 980 nm NIR laser, collimated and filtered before excitation. The specimen is mounted on a quartz plate atop an automated XY stage (Thorlabs NRT100 + LTS300), with scanning resolutions typically set to 200 × 100 points over areas of 4 × 2 cm (tissue) to 10 × 5 cm (whole mouse). For HSI, a monochromator with 150 lines/mm gratings disperses light collected via lenses, feeding into a liquid nitrogen-cooled InGaAs detector (320×256 pixels, 900–1700 nm). HDI bypasses the grating, using bandpass filters and a wide-field SWIR lens. A silicon CCD camera collects white-field images via beam-splitter for anatomical co-registration. The camera records the anatomy’s image which is co-registered to the spectrum of the anatomy. With this microscope, bright-field imaging increases the resolution of the specimen surface, allowing the development of a 3D mesh model of the specimen which is a scaffold for tissue fluorescence signal localization.

Probe Spectra

This system features both Hyperspectral Imaging (HSI) and Hyperdiffuse Imaging (HDI) modes with a trans-illumination configuration. A computer-controlled 980 nm near-infrared (NIR) laser provides illumination, which is collimated and filtered prior to the excitation. The specimen is held on a quartz plate mounted on an automated XY stage with scanning resolution at 200×100 points for an area of 4×2 cm to 10×5 cm. HSI uses a monochromator that disperses light with 150 lines/mm gratings, sending the light to a liquid nitrogen-cooled InGaAs detector. HDI bypasses the grating using bandpass filters and a wide-field SWIR lens. White-field images are collected with a silicon CCD camera via beam-splitter. The camera captures anatomical images, which are co-registered with spectra. Bright-field imaging increases resolution, and volumetric meshes are constructed as scaffolds for precise signal localization.

Significance

The DOLPHIN platform developed a new form of imaging that combines hyperspectral imaging with hyperdiffuse imaging. This system has enhanced capabilities for decomposing fluorescent signals as well as spatial mapping of photon diffusion, thus improving depth of imaging as well as sensitivity and resolution.

Principal Component Analysis

Band Selection

Principal component analysis (PCA) helps HSI/HDI data within 3D serve biological imaging by isolating spectral and spatial contrast features hidden by autofluorescence and scattered light. In HSI, the 4D data I(x, y, a, b) is transformed to I(x, y, lambda) by summing over the b dimension, where (x, y) are tissue coordinates and (a, b) are sensor pixels. (lambda) is wavelength. The data becomes a matrix ( x,y ... imes lambda ), and SVD is used to derive principal components(PC).

Each PC provides a weighted sum over ( lambda ), where high weights indicate importance. Selected bands include: - α-band near 980 nm (PC5), - β-band near 1100 nm (PC1,2,4), - γ-band near 1350 nm (PC3,5), - δ-band around 1600 nm (PC2).

In HDI, I(x, y, a, b) becomes I(x, y, r) , with r as radial distance from the sensor center. PCA is performed but final images use equal weights across radii.

Metrics Calculation

2D images are based on spectral intensity (SI) for HSI and diffuse intensity (DI) for HDI:

SI_{i = α−δ}(x, y) = ∑_λ HSC_{i = α−δ}(x, y, λ(i))

DI_{i = α−δ}(x, y) = ∑_r HDC_{i = α−δ}(x, y, r)

To improve contrast, SI ratios are calculated as:

SI_{i/j} = SI_{i = α−δ}(x, y) / SI_{j = α−δ, j ≠ i}(x, y)

Additional metrics: - **Spectral Position (SP):** λ where 50% of cumulative intensity is reached. - **Spectral Width (SW):** \( λ_{85} - λ_{15} \), the range covering 70% of intensity.

These help infer depth in 3D reconstructions.

Expected Results

DOLPHIN shows unmatched sensitivity and depth: - Up to 8 cm in breast phantoms, 6 cm in muscle, 2 cm in brain. - Detects as few as 208 cells at 2 cm depth. - ~100 µm clusters (200–1000 cells) can be imaged. - 200×100 resolution images at 20 ms/pixel over 10×5 cm in one scan.

"MIT" and Rat Results

In the MIT patterning test, three overlapping nanoparticles were labeled “M”, “I”, and “T”. - **M**: Er-NP (δ-band, 1575 nm) - **I**: Pr-NP (γ-band, 1350 nm) - **T**: Ho-NP (β-band, 1175 nm)

PCA successfully separated them using: - PC2 → δ-band - PC3 & PC5 → γ-band - PC1, PC2, PC4 → β-band

Spectral ratioing further improved separation.

In a rat experiment, a 1 mm Er-NP probe was located through a 4 cm thick specimen. Basic epi-illumination failed to detect it. However, DOLPHIN HDI localized the probe with Scattering Radius (SR) metrics using a Gaussian radial model. Result confirmed accurate spatial localization of sub-threshold fluorophores, demonstrating the power of HSI for labeling and HDI for 3D positioning.